2026 Optimizely Analytics release notes
6 min
Subscribe to updates on this page using the notification and RSS feed options.
September
- One-click experiment readouts – Turn an experiment's results into a readout dashboard with one click. Click Generate Readout on the new Readout tab of the experiment results page to build a results summary, the scorecard, and a time-series chart for each metric. Figures are frozen at generation, and the header shows when the readout was generated and the date its data runs through. Click Regenerate to rebuild it with the latest results.
- Experiment Diagnostics tab – Confirm that an experiment is reaching users and firing conversions from the new Diagnostics tab on the experiment results page. Switch between decision and conversion events, filter them, and click a row to see the raw event detail. (Not shown on public link views.)
- Feature rollouts in experiment scorecards – Analyze Feature Experimentation rollouts alongside A/B tests. Rollouts appear in the experiment selector labeled Feature Rollout, and traffic held out of a rollout appears as an Off baseline variation you can compare against. A rollout at 100% has no control group, so its scorecard reports only on the exposed population.
- Conversions Over Time chart – See how conversions built up over an experiment. Select the Conversions tab on a metric's chart to plot raw event counts per variation. The chart uses the scorecard's date range and conversion window. (Not available for every metric type.)
- Click-through drilldowns – Choose what happens when someone left-clicks a data point on a dashboard tile. Pick one action per tile in the visualization editor's Click-through drilldown tab. Actions include drilling down on the point or series, segmenting by cohort or attribute, and opening a saved exploration. (Single Value charts need the sparkline turned on.)
- Per-app query timeout – Give a slow warehouse more headroom on one application without raising the limit for every other app. Set Query Timeout (seconds) on the General Settings page, up to one hour.
- Sample Size Calculator – Plan an experiment before you run it. Set the audience, metric, traffic allocation, duration, and statistical thresholds. One exploration then shows the power curve, detectable lift, and estimated sample size. Power analyses save, reopen, and refresh like any other exploration. (Available per application. Turn on Sample Size Calculator under Feature Flags on the General Settings page.)
- Per-metric CUPED control – Turn CUPED on or off for any experiment metric, including guardrails. CUPED still applies by default to the first three eligible primary metrics, and the pre-experiment window defaults to 14 days.
- Clearer experiment summary – Read winners, losers, and guardrail alerts straight from the Summary table on the experiment results page. The table uses the same colors and outlines as the detailed scorecard. Values, improvement, and significance match each metric card, including when CUPED is on.
- Drag metrics between primary and guardrail sections – Drag experiment metrics to reorder them within the primary and guardrail sections or move them between the two. The drag handle also works from the keyboard.
- Key-pair authentication for Snowflake – New Snowflake connections default to Key-Pair Authentication, and setup no longer offers Simple Authentication (username and password). Existing password connections show a banner asking you to switch authentication modes to avoid disruption.
- OAuth connections for Databricks – Connect Databricks through each user's own Databricks account instead of a shared personal access token (PAT). Select OAuth U2M from the Authentication drop-down list on a Databricks connection, and each user clicks Connect to log in. Queries then run with that user's Databricks permissions, and a service account keeps scheduled reports and background jobs running.
- Recurring targets – Set a chart target to repeat daily, weekly, monthly, quarterly, yearly, or on a custom interval, so the deadline rolls forward instead of staying overdue.
- Cancel a running exploration query – Canceling an exploration stops the query in the warehouse, not just in the browser. Analytics discards any late result instead of showing it.
- Public link from embedded experiment results – Create a public link from the header of embedded experiment results, beside Help. The option is available to users with full access to the experiment.
- Create cohorts with Mark – Describe the users you want, and Mark (Optimizely Agent Platform) saves them as a reusable cohort. Ask for a behavioral cohort based on events, event sequences, or user properties, or a retention cohort of users who returned or churned after a starting event. Behavioral cohorts can nest AND, OR, and NOT conditions up to three levels deep, and retention cohorts can filter the return event by property. Mark can also define a cohort on one dataset and reduce it to a related one, such as users whose sessions include a content play more than three times.
August
- Public URLs for experiment results – Share experiment results the way you already share dashboards. Generate a public link or an embeddable iframe from the Share modal. Viewers open a read-only results page without an Optimizely Analytics login.
- Statistical significance notifications – Opt in per experiment to be notified when a metric reaches statistical significance, so you can act without watching the scorecard. Sequential and Bayesian experiments are supported, with one notification per metric and variation combination.
- Custom color palettes – Set an app-wide color palette so every chart matches your brand. Choose from predefined palettes or build and save your own.
- Event usage stats in the event selector – See how heavily each event type is used while you build an analysis. The selector shows an average daily count, color-coded so busier events read darker, which makes stale events easy to spot.
- Value counts in column details and filters – Column and JavaScript Object Notation (JSON) field tooltips show the top values, distinct counts, and null percentage, so you can see what a column holds before filtering on it. Distinct-value dropdowns list each value's occurrence count and its share of the column.
- Conversion drivers in Mark – Ask Mark what drove or blocked conversion between funnel steps. Mark returns a link to the conversion driver analysis, so you no longer have to build it manually. Supports single segment analyses.
- Mark dashboard actions – Run Mark actions on a dashboard from one menu: Summarize Dashboard, Arrange Tiles, and Revert Layout. New Ask Mark entry points in the sidebar and type picker start an exploration without leaving your dashboard.
July
- Unified experiment results page – Added a redesigned experiment results page. A guided tour walks you through reading and interacting with the page.
- Embedded dashboards – Embed full dashboards in external sites and tools, so a team reads your dashboard where they already work. Embed URLs include readable parameter and filter names. You can embed visualizations directly from the Share modal through a new Embedding tab.
- Drag-and-drop dashboard tiles – Reorder and resize dashboard tiles by dragging them, so you can rearrange a dashboard without editing each tile. Grid units appear while resizing, and new tiles land full-width at the bottom.
- First-ever event option for funnels – Count the first funnel stage only when the event is the actor's first-ever occurrence, not just the first within the time range. This isolates new behavior.
- Event occurrence filter for event segmentation – In event segmentation with a Uniques measure, choose how occurrences count: all matching events in the time range, only the first in range, or each actor's first-ever occurrence. This separates new behavior from repeat behavior.
- Inline description editing in explorations – Edit event and custom column descriptions in place from the exploration UI, saved from the tooltip and details views. Context stays current without a trip to the catalog.
- Nested JSON fields in the dataset editor – Drill into JSON-typed columns in the Columns tab to curate their sub-fields inline, so nested data is as discoverable as a top-level column. Enable or disable fields, set data index, assign categories, and edit descriptions.
- Inline actor lookup – Inspect the individual actors behind a chart without leaving the visualization. A paginated actor table opens inline on a tile, and clicking a row jumps to that actor's details.
- Mark workspace overview – Mark now checks which datasets, events, and columns exist before it answers, so its analyses reflect your real workspace instead of assumed structure.
- Mark Structured Query Language (SQL) tool – Mark runs read-only SELECT queries against the warehouse as a last resort when no existing explore type answers the question, returning rows with their schema.
- Automatic experiment pause on a guardrail breach – Opt in per guardrail alert to automatically pause an experiment when a metric breaches its threshold, so a degraded experience stops without waiting for manual intervention.
- Time range filters on scorecards – Apply a dashboard time range filter to narrow an experiment scorecard's analysis window after the fact. You can use it to exclude a day affected by an unrelated site incident.
June
- Conversion drivers – Understand what users did between funnel steps with a beeline analysis that maps the behaviors driving or blocking conversion, surfacing the events that best explain drop-off after any step.
- Automatic sample ratio mismatch check – Sample ratio mismatch (SRM) is checked automatically when the experiment results page opens or refreshes, so you no longer need the manual refresh button. Clear Not available states explain when a check is skipped.
- Progressive dashboard filters – Link dashboard filter tiles together so one filter's dropdown narrows based on the selection in another. Filtered dashboards become faster to navigate.
- Public dashboard sharing – Share dashboards with password-protected links that expire on a date you set. Executives and external partners then view read-only dashboards without an Optimizely Analytics login. Link management includes per-link enable and disable, bulk revoke, and a per-app control to allow or block public link creation.
- Background query refresh – Experiment and exploration results load immediately from cache while a refresh runs in the background, replacing the blank screen you used to see while waiting on recalculation.
- Bulk delete events – Select multiple events in the event catalog and delete them in one action, so cleaning up a cluttered catalog no longer means deleting events one at a time. A confirmation dialog shows how many you are removing.
- Categories and favorites in global search – Global search results show the categories applied to an entity and a star for favorited entities, so you can tell items apart without opening each one.
- Mark engagement analysis – A new engagement tool lets Mark analyze event engagement before and after a focal point. For example, compare feature usage two weeks either side of a product launch.
- Mark event measure types – Build analyses through Mark using four more event segmentation measure types: Frequency, Active Actor Percent, Intervals Engaged, and Custom Property with percentile aggregation. Mark now covers the measure types you already use in the UI.
May
- Optimizely Analytics Model Context Protocol (MCP) server – Connect your AI tools to Optimizely Analytics through the Analytics MCP server, so an agent queries your data and builds analyses without leaving the tool you already work in.
- Funnel metrics – Measure sequential event completion as a first-class metric type. Define an ordered sequence of steps and track conversion rates between stages. Funnel measures work directly in experiment scorecards, so you can see how a treatment affects a multi-step journey.
- Automatic CUPED selection – Controlled-experiment Using Pre-Experiment Data (CUPED) evaluates results with and without variance reduction in parallel and applies it only when it improves the statistics. Each metric shows whether CUPED was used and why.
- Chart annotations – Add notes directly on time-series charts to capture context around key moments, so the story behind your data travels with the chart.
- Exploration targets – Set a numeric target on any measure and see at a glance whether you are on track. Single value tiles show a progress bar with percent-to-goal, and line or bar chart tiles show a dashed target line.
- Single value period-over-period comparison – Headline chart tiles display period-over-period change alongside the main metric. The delta value, percentage change, and a directional color indicator show trend direction at a glance.
- Audience conditions in experiment results – The Audiences field on the experiment results page displays the audience conditions applied to the experiment in a structured tree, giving you clearer targeting context alongside your metrics.
- Experiment conclusions – Concluded experiments surface their conclusion, deployed variation, and conclusion date directly on the results page, so decision context stays visible without digging through history.
- Refresh experiment results – Fetch the latest experiment results on demand with Refresh results in the results toolbar, without reloading the full page.
- User visibility controls – Organization admins can toggle whether users see all members of the organization or only those sharing a group, so you can keep membership private in shared environments.
- Case-sensitive contains filter – Configure each column individually for case-sensitive CONTAINS comparisons, which can speed up filtering on warehouses where the default case-insensitive match is expensive.
- Right-click legend highlight – Right-click a series in a chart legend to hide all the others, so you can read one series in a crowded chart. Right-click again to restore full visibility. Works for bar, line, pie, and doughnut charts.
- Persistent column widths – Column width adjustments in data tables are saved, so your layout preferences persist across sessions.
- Event selector usage sorting – Sort the event selector by your own and your team's usage, so the events you use most display first. Switch to alphabetical order anytime, and Analytics saves your choice for each app.
- Mark experiment and dashboard tools – Mark builds experiment scorecards with specific metrics, adds analyses to an experiment's Explore tab, and adds image tiles to dashboards. Together these enable an end-to-end automated analysis workflow.
- Mark comma-separated values (CSV) upload – Upload CSV data directly to your data warehouse from Mark, using inline CSV content or an authenticated file URL, up to 20 MB.
- Mark nested JSON support – Mark tools reference nested JSON column properties in filters, group-bys, and measures. Previously only top-level column IDs were accessible.
April
- CUPED variable configuration – Control variance reduction precisely in experiment scorecards. The CUPED settings let you select specific experiment measures and actor-level properties as CUPED variables, replacing the previous on and off toggle.
- Experiment results page redesign – Added a redesigned experiment results experience. It adds an experiment time range display, visitor counts, and clearer empty states when data conditions are not yet met.
- Edit explore while querying – Continue refining your exploration configuration while a query is in flight. Changes are preserved without interrupting the running query.
- Mark default app preference – Set a preferred app for Mark per user and per organization. Mark resolves to your preferred app without a manual selection step each session.
- Historical event counts timeline – The organization data usage page loads a time series of historical event counts automatically, so you see ingestion trends without clicking Check data usage.
- Duplicate measure blocks – Duplicate a measure in a universal exploration to build a variation without re-creating its configuration.
- Formula editor hover cards – Hover over @ mentions in the formula editor to see entity details without leaving your formula. Slash and quoted names are also valid in variable references.
- Color picker reset – Click Reset to default to return a customized series color to its original, so you can undo a manual color change without rebuilding the chart.
- Events category assignment – Assign raw and clean events to categories directly from the Events page, so you can organize a new event without opening the catalog.
- Collapsible editor pane – Click Hide Editor on the experiment results page to collapse the editor and free up horizontal space when you review results. Click Expand Editor to bring it back.
March
- Influence exploration template – Identify which user behaviors, events, and metrics are most strongly associated with key business outcomes such as conversions, revenue, and retention.
- Image tile in dashboards – Add static images to Optimizely Analytics dashboards, such as experiment variation screenshots or brand assets, directly alongside data and metrics.
- Bounce and exit rate metrics – Measure user engagement and pinpoint friction by tracking bounces in a given session and identifying the exact pages where visitors leave your site (exits).
- SCIM for admin center – Streamline user management and enhance enterprise security by automating the provisioning and de-provisioning of user accounts directly through the Admin Center.
- Released the following Analytics system tools in Mark to help you create explorations from Mark:
- oa_add_explore_to_dashboard – Pin your saved data explorations directly to any of your analytics dashboards as a visual tile.
- oa_create_dashboard – Instantly generate custom dashboards to start organizing your most important explorations in one place.
- oa_add_text_tile_to_dashboard – Enhance your dashboards by adding customizable text headers and descriptions to provide clear narrative context for your data.
- oa_analyze_experiment – Automatically fetch and interpret the scorecard results of an experiment to understand variation performance and statistical significance.
- oa_analyze_experiment_explore_tab – Discover and access the underlying custom explorations that drive the deeper insights within an experiment's Explore tab.
- oa_arrange_dashboard_tiles – Seamlessly organize your dashboard layout by repositioning and resizing your data visualizations on a flexible grid.
- oa_revert_dashboard_layout – Instantly undo your most recent dashboard layout changes to safely restore the previous tile arrangement.
February
- Explore results dropdown in experiment summary – Access detailed result views directly from the experiment summary section through a Explore results drop-down.
- Released the following Analytics system tools in Mark to help you create explorations from Mark:
- oa_find_metrics – Searches and finds existing metrics from Optimizely Analytics. Uses fuzzy matching on metric names and semantic search on descriptions to find relevant metrics.
- oa_get_metric_data – Retrieves and executes a specific saved metric in Optimizely Analytics, returning its calculated table data.
- First time ever filter – Filter Event Segmentation explorations to track unique users taking a specific action for the first time ever.
January
- Series-level color labeling – Customize the color and label for individual data series in event segmentation charts to create more consistent and readable dashboards.
- Updated formula block editor - Streamline formula creation with a more reliable autocomplete and better search functionality in the formula editor, making it faster to build complex calculated fields.
- Data integrity health checks – Validate experiment data quality including visitor ID consistency, assignment overlap between datasets , and primary key uniqueness to identify and resolve data issues impacting experiment validity.
- Mark knowledge expansion – Get more contextual and relevant responses to your analytics questions. Mark now accesses and analyzes your existing explorations and metrics, letting you generate new analyses with greater precision.
- Chart legend improvements – Get enhanced legend rendering for improved performance and readability, especially with high-cardinality data. You can now scroll through legends and view all items without truncation.